TechnischeUniversitatChemnitz-Zwickau



Similar documents
Schneps, Leila; Colmez, Coralie. Math on Trial : How Numbers Get Used and Abused in the Courtroom. New York, NY, USA: Basic Books, p i.

SPLIT PLOT DESIGN 2 A 2 B 1 B 1 A 1 B 1 A 2 B 2 A 1 B 1 A 2 A 2 B 2 A 2 B 1 A 1 B. Mathematical Model - Split Plot


Torgerson s Classical MDS derivation: 1: Determining Coordinates from Euclidean Distances

SOLVING LINEAR SYSTEMS

On the Algebraic Structures of Soft Sets in Logic

An Intersection Theorem for Systems of Sets

Direct Methods for Solving Linear Systems. Matrix Factorization

4. CLASSES OF RINGS 4.1. Classes of Rings class operator A-closed Example 1: product Example 2:

Factorization Theorems

KeyEscrowinMutuallyMistrustingDomains?

TECHNIQUES OF. C.T. Pan 1. C.T. Pan

3. INNER PRODUCT SPACES

Iterative Solvers for Linear Systems

OPTIMAL SELECTION BASED ON RELATIVE RANK* (the "Secretary Problem")

How To Prove The Dirichlet Unit Theorem

Solution of Linear Systems


it is easy to see that α = a

The Assignment Problem and the Hungarian Method

= y y 0. = z z 0. (a) Find a parametric vector equation for L. (b) Find parametric (scalar) equations for L.

Reference: Introduction to Partial Differential Equations by G. Folland, 1995, Chap. 3.

LECTURE 4. Last time: Lecture outline

Tail inequalities for order statistics of log-concave vectors and applications

Using a Balanced Scorecard to Tie the Results Act to Your Day-to-Day Operational Priorities

On Quantum Hamming Bound

PROBLEM SET 6: POLYNOMIALS

Solution to Homework 2

ARTICLE IN PRESS. International Journal of Lean Thinking Volume 1, Issue 1 (June 2010)

Limits and Continuity

Some generalized fixed point results in a b-metric space and application to matrix equations

A Robust Optimization Approach to Supply Chain Management

1 Determinants and the Solvability of Linear Systems

Let H and J be as in the above lemma. The result of the lemma shows that the integral

15. Symmetric polynomials

Lecture Notes on The Mechanics of Elastic Solids

Modern Algebra Lecture Notes: Rings and fields set 4 (Revision 2)

CS473 - Algorithms I

Approximability of Two-Machine No-Wait Flowshop Scheduling with Availability Constraints

Scheduling a sequence of tasks with general completion costs

On the irreducibility of certain polynomials with coefficients as products of terms in an arithmetic progression

Abstract: We describe the beautiful LU factorization of a square matrix (or how to write Gaussian elimination in terms of matrix multiplication).

Linear Threshold Units


SF2940: Probability theory Lecture 8: Multivariate Normal Distribution

THE EXISTENCE OF CONVEX BODY WITH PRESCRIBED CURVATURE MEASURES. 1. Introduction

Figure2:Themixtureoffactoranalysisgenerativemodel. j;j z

9.2 Summation Notation

Equilibrium computation: Part 1

Control Systems with Actuator Saturation

I. GROUPS: BASIC DEFINITIONS AND EXAMPLES

A linear algebraic method for pricing temporary life annuities

1 Review of Newton Polynomials

FUZZY CLUSTERING ANALYSIS OF DATA MINING: APPLICATION TO AN ACCIDENT MINING SYSTEM

THE GENERALIZED RECURRENT WEYL SPACES HAVING A DECOMPOSABLE CURVATURE TENSOR

MATRIX ALGEBRA AND SYSTEMS OF EQUATIONS. + + x 2. x n. a 11 a 12 a 1n b 1 a 21 a 22 a 2n b 2 a 31 a 32 a 3n b 3. a m1 a m2 a mn b m

VERTICES OF GIVEN DEGREE IN SERIES-PARALLEL GRAPHS

AP Calculus BC Exam. The Calculus BC Exam. At a Glance. Section I. SECTION I: Multiple-Choice Questions. Instructions. About Guessing.

NOV /II. 1. Let f(z) = sin z, z C. Then f(z) : 3. Let the sequence {a n } be given. (A) is bounded in the complex plane

TITLE 9. HEALTH SERVICES CHAPTER 1. DEPARTMENT OF HEALTH SERVICES ADMINISTRATION ARTICLE 4. CODES AND STANDARDS REFERENCED

LEARNING OBJECTIVES FOR THIS CHAPTER

LIBRARY OF THE MASSACHUSETTS INSTITUTE OF TECHNOLOGY

UNIVERSITETET I OSLO

1 Completeness of a Set of Eigenfunctions. Lecturer: Naoki Saito Scribe: Alexander Sheynis/Allen Xue. May 3, The Neumann Boundary Condition

LIST OF RANK AND FILE CLASSES IN BU 9 As of December 1, 2011

MAT188H1S Lec0101 Burbulla

Irreducibility criteria for compositions and multiplicative convolutions of polynomials with integer coefficients

SF2940: Probability theory Lecture 8: Multivariate Normal Distribution

Matrix-Chain Multiplication

Tutorial: Operations Research in Constraint Programming

ON LIMIT LAWS FOR CENTRAL ORDER STATISTICS UNDER POWER NORMALIZATION. E. I. Pancheva, A. Gacovska-Barandovska

The Characteristic Polynomial

Chapter 4, Arithmetic in F [x] Polynomial arithmetic and the division algorithm.

Module MA3411: Abstract Algebra Galois Theory Appendix Michaelmas Term 2013

Pain-Free Injections

The Australian Journal of Mathematical Analysis and Applications

The Matrix Elements of a 3 3 Orthogonal Matrix Revisited

Lecture 2 Matrix Operations

Econ 430 Lecture 9: Games on Networks

1 = (a 0 + b 0 α) (a m 1 + b m 1 α) 2. for certain elements a 0,..., a m 1, b 0,..., b m 1 of F. Multiplying out, we obtain

WHAT DEGREE OF CONFLICT IN THE DOMAIN OF INDIVIDUAL PREFERENCES COULD RESOLVE THE PARADOX OF DEMOCRATIC MAJORITY DECISION?

Transcription:

TechnischeUniversitatChemnitz-Zwickau DFG-Forschergruppe\SPC"FakultatfurMathematik DomainDecompositionand PreconditioningOperators MultilevelTechniquesfor SergejV.Nepomnyaschikh SiberianBranchofRussianAcademyofSciences e-mail:svnep@comcen.nsk.su ComputingCenter Novosibirsk,63090 SPC9530 Preprint-ReihederChemnitzerDFG-Forschergruppe \ScienticParallelComputing" November1995

1Introduction Inrecentyears,domaindecompositionmethodshavebeenusedextensivelytoeciently 5,15,17].Butdirectrealizationofmultileveltechniquesonaparallelcomputersystem havedevelopedintoaneectivetoolfortheconstructionandanalysisoffastsolvers[2, fortheglobalproblemintheoriginaldomaininvolvesdicultcommunicationproblems. solveboundaryvalueproblemsforpartialdierentialequationsincomplex{formdomains [4,13,16].Ontheotherhand,multileveltechniquesonhierarchicaldatastructuresalso Ithispaper,wepresentandanalyzeacombinationofthesetwoapproaches:domain preconditioningoperators. decompositionandmultileveldecompositiononhierarchicalstructurestodesignoptimal LetR2beapolygon.Inthedomainweconsidertheboundaryvalueproblem 8><>:?2X where i;j=1@ @xiaij(x)@u @N=2X @u@xj+a0(x)u=f(x);x2 i;j=1aij(x)@u @N+(x)u=0; u(x)=0; @xjcos(n;xi) x2?1: x2?0 (1.1) of?0: Weintroducethebilinearforma(u;v)andthelinearfunctionall(v): nitenumberofcurvilinearsegments,?=?0[?1;?0=?0:here?0denotestheclosure istheconormalderivative,ndenotestheoutwardnormalto?,and?0isaunionofa ByH1(;?0)wedenotethesubspaceoftheSobolevspaceH1() a(u;v)=z2x H1(;?0)=nv2H1()jv(x)=0;x2?0o: (1.1)aresuchthatthebilinearforma(u;v)issymmetric,elliptic,andcontinuouson Letussupposethattheoperatorcoecientsandtheright-handsideoftheproblem i;j=1aij(x)@u l(v)=zf(x)vdx: @xj@v @xi+a0(x)uvdx+z?1(x)uvdx H1(;?0)H1(;?0),i.eȧ(u;v)=a(v;u)8u;v2H1(;?0) andthelinearfunctionall(v)iscontinuousonh1(;?0): tionproblem[1]u2h1(;?0):a(u;v)=l(v)8v2h1(;?0): Thegeneralizedsolutionu2H1(;?0)of(1.1)is,bydenition,asolutiontotheprojec- 0kuk2H1()a(u;u)1kuk2H1()8u2H1(;?0) jl(u)jkukh1()8u2h1(;?0): 1 (1.2)

Weknowthatundertheseassumptionsfora(u;v)andl(v)thereexistsauniquesolution of(1.2). whereiarepolygonswithdiametersontheorderofh.letusconsideracoarsegrid triangulationof Letbeaunionofnnonoverlappingsubdomainsi, =n[i=1i;i\j=;;i6=j; andwerenehi;0severaltimes.thisresultsinasequenceofnestedtriangulations h0=n[i=1hi;0;hi;0=m(0) diam((0) hi;0;hi;1;:::;hi;j i;l)=0(h) i[l=1(0) i;l; suchthat wherethetriangles(k+1) subtrianglesbyconnectingthemidpointsoftheedges. IntroducethespacesWi;0Wi;1:::Wi;J=Hh(i); i;laregeneratedbysubdividingtriangles(k) hi;k=m(k) i[l=1(k) i;l;k=0;1;:::;j; Vi;0Vi;1:::Vi;J=Hh(?i); i;lintofourcongruent onthetrianglesinhi;k.thespacevi;kisthespaceoftraceson?ioffunctionsfromwi;k: HerethespaceWi;kconsistsofreal-valuedfunctionswhicharecontinuousonandlinear Vi;k=n'hj'h=uhj?i;withuh2Wi;ko:?i=@i;i=1;2;:::;n: (1.3) whichisanapproximationoftheproblem(1.2). WedenethespaceHh()ofrealcontinuousfunctionswhicharelinearoneachtriangle ofhandvanishat?0: triangulationh.then(1.4)isequivalenttothesystemofmeshequations RNwhosecomponentsarevaluesofthefunctionuhatthecorrespondingnodesofthe Letusconsidertheprojectionproblem Eachfunctionuh2Hh()isputincorrespondencewitharealcolumnvectoru2 uh2hh():a(uh;vh)=l(vh)8vh2hh() Au=f; (1.4) (Au;v)=a(uh;vh)8uh;vh2Hh(); (f;v)=l(vh)8vh2hh(); 2 (1.5)

whereuhandvharetherespectiveinterpolationsofvectorsuandv;(f;v)istheeuclidean scalarproductinrn: Thegoalofthisworkistoconstructasymmetricpositivedenitepreconditioning operatorbfor(1.5)soastosatisfytheinequalities c1(bu;u)(au;u)c2(bu;u) (1.6) wherethepositiveconstantsc1andc2areindependentofhandh,themultiplicationof avectorbyb?1shouldbeeasytoimplement. UsingacombinationofAdditiveSchwarzandFictitiousSpaceMethods,optimalpreconditioningoperatorshavebeenconstructedin[11,12,13]forthecaseofarbitrary (unstructured)grids.however,thatconstructioninvolvesexplicitextensionoperators whoseimplementationforthreedimensionalproblemsisoptimalfromthearithmeticcost andtheconditionnumberpointsofviewbutdicultforpracticerealization.themain goalofthisworkistoconstruct,usingthehierarchicalstructure(1.3),arobustoptimal preconditioningoperator.oneofthecrucialpointsin[11,12,13]andthispaperisusingof non{exactsolversinsubdomainsandexplicitextensionoperators.itmeans,toconstruct optimalpreconditioningoperators,wecandesignnormpreservingoperatorsoffunctions givenat?iintoiwiththeoptimalarithmeticcost(anumberofarithmeticoperations shouldbeproportionaltoanumberdegreesoffreedom)andthen,insteadofexactsolvers insubdomains,wecanuseanyspectrallyequivalentpreconditioningoperators.optimal extensionoperatorshavebeenpresentedin[8,9,11]forunstructuredgridsandrobust explicitextensionoperatorsonhierarchicaldatastructuresin[5,14]. Thepaperisorganizedasfollows.IntheSection2,usingAdditiveSchwarzMethod, wedescribegeneralconstructionofapreconditioningoperatorwithlocalmultilevelpreconditioningoperators.inthesection3,wepresentanoptimalmultilevelextensionof gridfunctionsfromboundariessubdomainsintoinsidesubdomains.inthesection4, weproposeanoptimalinterfacepreconditioningoperatorattheboundariesofthesubdomainswhichinvolvesamultileveldecompositionandcorrespondingexplicitextension operatorsatinterfaces. 2Domaindecomposition{additiveSchwarz-Method Todesignthepreconditioningoperatorforthesystem(1.5),weusetheadditiveSchwarz{ Method[7]andrealizethemainideaoftheconstructionofpreconditionersfrom[13]for thehierarchicalgrids.denotebyhh(i)thesubspaceofhh(i) Hh(i)=nuh2Hh(i)juh(x)=0;x2?io anddenethelocalpreconditioningoperatorsbisuchthat Bi:Hh(i)!Hh(i); c3kuhk2h1(i)(biu;u)c4kuhk2h1(i)8uh2hh(i); wherec3;c4areindependentofhandh.wehereafterusethesamedesignationforan operatoranditsmatrixrepresentation.forinstance,todenebi,wecanusetheso-called BPX{preconditioners[3].Todoit,denotebyff(k) lgnodalbasisfunctionsfromthek{th levelanddene B?1 iuh=jxk=0x f(k) l2hh(i)(uh;f(k) l)l2(i)f(k) l: (2.1) 3

suchthat Letusassumethatwecandenetheextensionoperatorsti uh(x)='h(x);x2?i; ti'h=uh; ti:vi;j?!wi;j withc5independentofhandh.herek'hkh1=2(?i)isthenorm[10]inthesobolevspace H1=2(?i)k'hk2H1=2(?i)=HZ?i('h(x))2dx+Z?iZ?i('h(x)?'h(y))2 kti'hkh1(i)c5k'hkh1=2(?i)8'h2vi;j; jx?yj2dxdy: (2.2) Then,wecandenetheextensionoperatort andforany'h2hh(s) wherehh(s)isthespaceoftracesoffunctionsfromhh()ats t:hh(s)!hh(); uh(x)='h(x);x2s; t'h=uh; S=n[i=1?i Here Theoperatortifrom(2.2)isconstructedintheSection3. Letsatisestothefollowinginequalities kt'hkh1()c5k'hkh1=2(s): c6k'hk2h1=2(s)(';')c7k'hk2h1=2(s)8'h2hh(s); k'hk2h1=2(s)=nxi=1k'hk2h1=2(?i): ditioningoperatorbasfollows wherec6;c7independentofhandh.then,accordingto[11],wecandenetheprecon- B?1=2640B?1 1...B?1 n375+t?1t: (2.3) Here0isthenull-matrixwhichcorrespondstonodesofthetriangulationhatSandBi isfrom(2.1). Thefollowingtheoremisvalid (2.4) independentofhandh. Theorem2.1IftheoperatorBisfrom(2.4),thentheconstantsc1;c2from(1.6)are 4

Themaingoalofthissectionistoconstructtherobustoperatortifrom(2.2).During thissection,weomitthesubscripti. 3Multilevelexplicitextensionoperators wefollowto[5,14].denoteby'(k) Todesigntheextensionoperatort:VJ!WJ; thel2-orthoprojectionfroml2(?)onto(k) itheone-dimensionalsubspacespannedbythisfunction'(k) ~Qk=Nk i;i=1;2;:::;nk,thenodalbasisofvkanddenoteby Xi=1Q(k) i:l2(?)!(k) i;k=0;1;:::;j?1: ianddenote i i:dene any'h2vjk'h0k2h1=2(?)+1hk'h1k2l2(?)+j'h1j2h1=2(?)c8k'hk2h1=2(?); Thefollowinglemmasarevalid[14]. Lemma3.1Thereexistsapositiveconstantc8,independentofhandH,suchthatfor Fork=Jwedene~QJastheL2-orthoprojectionfromL2(?)ontoVj. Lemma3.2Thereexistsapositiveconstantc9,independentofhandH,suchthat where Here j'hj2h1=2(?)=z?z?('h(x)?'h(y))2 'h0=~q0'h;'h1='h?'h0: jx?yj2dxdy: (3.1) Denotebyx(k) where'h0;'h1from(3.1). TheconstructionoftheoperatortisbasedonthedecompositionfromtheLemma3.2. k'h0k2+1hk~q0'h1k2l2(?)+jxk=12kk(~qk?~qk?1)'h1k2l2(?)c9k'hk2h1=2(?); thefollowingway.forany'h2vjset iareenumeratedrston?andtheninside)anddenetheextensionoperatortin i;i=1;2;:::;lk,thenodesofthetriangulationhk(weassumethatnodes Then hk=(~qk?~qk?1)'h;k=1;2;:::;j: h0=~q0'h; 'h= h0+ h1+:::+ hj: (3.2) 5

Denetheextensionuhk2Wkasfollows uh0(x(0) ; h0(x(0) i);x(0) Hereis,forinstance,themeanvalueofthefunction uhk(x(k) i)=8<: k=1;2;:::;j: 0; hk(x(k) i);x(k) i=2?; i2?; i62?; h0on? (3.3) =1N0N0 Xi=1 0;1;:::;J?1:Butinthiscasethecostofthedecomposition(3.2)isexpensive(especially Dene Remark3.1WecanusetheL2-orthoprojectionsformL2(?)ontoVkinsteadof~Qk;k= t'h=uhuh0+uh1+:::+uhj h0(x(0) i): Theorem3.1Thereexistsapositiveconstantc10,independentofhandH,suchthat forthreedimensionalproblems). (3.4) Heretheoperatortisfrom(3.2){(3.4). Remark3.2Itisobviousthat kt'hkh1()c10k'hkh1=2(?)8'h2vj: andthecostoftheactionoftandtisproportionaltothenumberofnodesofthegrid domain. Q(k) i'h=('h;'(k) 4Interfacepreconditioningoperators ('(k) i;'(k) i)l2(?)'(k) i [13].LetSbeaunionofKnonoverlappingedgesEiofthetriangulationh0 satises(2.3).todoit,weusetheideaofadditiveschwarzmethodatinterfacesfrom Inthissection,weconstructanoptimalinterfacepreconditionerinthespaceHh(S)which SplitHh(S)intoavectorsumofsubspaces S=K[ whereu0isthecoarsespacewhichconsistsofcontinuousfunctionslinearontheedges Hh(S)=U0+U1+:::+UK; j=1ej;ej\ei=;;i6=j: Ej,j=1;2;:::;K,andUj,j=1;2;:::;K,correspondtoEjandaredenedbelow. 6 (4.1)

ForanyedgeEjwedenetheexplicitextensionoperatorj Denoteby Uj=f'h2Hh(S)j'h(x)=0;x62Ejg; ~U(k) aredierfromthefunctions'(k) asfollows.denoteby'(k) j=vkjej;k=0;1;:::;j: denoteby(k) j;i,i=1;2;:::;i(k) j:~u(j) j!hh(s) correspondingl2-orthoprojection.set j;itheone-dimensionalsubspacespannedbythisfunction'(k) ifromthesection3onlyattheendpointsofej)and j;i:l2(ej)!(k) j,thenodalbasisof~u(k) j(thefunctions'(k) j;i:denej;i extensionoperatorjaccordingto(3.2){(3.4).forany'h2~u(j) anddene~q(k) JastheL2-orthoprojectionfromL2(Ej)onto~U(J) ~Q(k) j=i(k) jxi=1q(k) j;i;k=0;1;:::;j?1; j:nowwecandenethe and hk=(~q(k) h0=~q(0) j'h; j?~q(k?1) j)'h;k=1;2;:::;j jset uhk=8><>: 0; hk(x(k) i);x(k) i2ej i62ej;k=0;1;:::;j (4.3) (4.2) theorem. Dene ThenfromtheTheorem3.1and[13]forthedecomposition(4.1)wehavethefollowing j'h=uh0+uh1+:::+uhj: Theorem4.1Thereexistsapositiveconstantc11;dependentofhandH,suchthatfor anyfunction'h2hh(s)thereexist'hj2uj,j=0;1;:::;k,suchthat Uj=Uj+j~Uj: Lettheoperator0generatesanequivalentnorminU0: k'h0k2h1=2(s)+k'h1k2h1=2(s)+:::+k'hkk2h1=2(s)c11k'hk2h1=2(s) c12k'hk2h1=2(s)(0';')c13k'hk2h1=2(s)8'h2u0; 'h0+'h1+:::+'hk='h; 7 (4.4)

DenotebyjandejtheBPX-likepreconditionersinthespacesUjand~Uj,respectively wherec12;c13independentofhandh.denelocalpreconditionersforuj,j=1;2;:::;k. Then,denetheinterfacepreconditioningoperatorinthefollowingway?1 i'h=jxk=0x supp'(k) j;iej('h;'(k) j;i\ej6=;('h;'(k) j;i)l2(ej)'(k) j;i8'h2uj;?1=+0+kxj=1(?1 j+je?1 jj): j;i8'h2~uj: operator?1 Here+0isapseudo-inverseof0from(4.4),jisfrom(4.2),(4.3),andweextendthe independentofhandh. Theorem4.2Iftheoperatorisfrom(4.5)thentheconstantsc6;c7form(2.3)are jbyzerooutsideej.thefollowingtheoremisvalid. (4.5) dimensionalproblems. Remark4.2Usingcombinationofpresentedtechniqueandtechniquefrom[10],eective Remark4.1Themethodsuggestedinthispapercanbegeneralizedevidentlyforthree TheauthorwishtoacknowledgetheDFG{Forschergruppe\SPC"ChemnitzandtheISF forsupportingthisresearchundergrantsla767/3andnpb000. preconditioningoperatorsforellipticproblemswithjumpcoecientscanbeconstructed. References Acknowledgment [1]J.-P.Aubin.Approximationofellipticboundary{valueproblems.Wiley{Interscience, [2]J.H.Bramble.MultigridMethods.ResearchNotesinMathematicsSeries.Pitman, NewYork,London,Sydney,Toronto,1972. [4]M.DryjaandO.B.Widlund.Multileveladditivemethodsforellipticniteelementproblems.InW.Hackbusch,editor,ParallelAlgorithmsforPartialDierential Comput.,55(191):1{22,1990. Boston-London-Melbourne,1993. [3]J.H.Bramble,J.E.Pasciak,andJ.Xu.Parallelmultilevelpreconditioners.Math. Equations,pages58{69,Braunschweig,1991.Vieweg{Verlag.Proc.oftheSixth [5]G.Haase,U.Langer,A.Meyer,andS.V.Nepomnyaschikh.Hierarchicalextension GAMM{Seminar,Kiel,January19{21,1990. andlocalmultigridmethodsindomaindecompositionpreconditioners.east{west J.Numer.Math.,2(3):173{193,1994.8

[7]A.M.MatsokinandS.V.Nepomnyaschikh.ASchwarzalternatingmethodina [6]W.Hackbusch.Multi{GridMethodsandApplications,volume4ofSpringerSeries [8]A.M.MatsokinandS.V.Nepomnyaschikh.Normsinthespaceoftracesofmesh incomputationalmathematics.springer{verlag,berlin,1985. subspace.sovietmathematics,29(10):78{84,1985. [10]S.V.Nepomnyaschikh.Domaindecompositionmethodsforellipticproblemswith [9]S.V.Nepomnyaschikh.DomaindecompositionandSchwarzmethodsinasubspacefor discontinuouscoecients.inr.glowinski,y.a.kuznetsov,g.a.meurant,and CenteroftheSiberianBranchoftheUSSRAcademyofSciences,Novosibirsk,1986. theapproximatesolutionofellipticboundaryvalueproblems.phdthesis,computing functions.sov.j.numer.anal.math.modelling,3:199{216,1988. [11]S.V.Nepomnyaschikh.Methodofsplittingintosubspacesforsolvingellipticboundaryvalueproblemsincomplex{formdomains.Sov.J.Numer.Anal.Math.Modelling, 6:151{168,1991. andtheirinversion.sov.j.numer.anal.math.modelling,6:1{25,1991. J.Periaux,editors,Domaindecompositionmethodsforpartialdierentialequations, pages242{251,philadelphia,1991.siam.proceedingsofthe4thinternationalsymposium,moscow,1990. [13]S.V.Nepomnyaschikh.DecompositionandFictitiousDomainsMethodsforElliptic [12]S.V.Nepomnyaschikh.Meshtheoremsontraces,normalizationoffunctiontraces [15]P.Oswald.MultilevelFiniteElementApproximation:TheoryandApplications. [14]S.V.Nepomnyaschikh.Optimalmultilevelextensionoperators.PreprintSPC953, BoundaryValueProblems.InT.F.Chan,D.E.Keyes,G.A.Meurant,T.S.Scroggs, TeubnerSkriptenzurNumerik.B.G.TeubnerStuttgart,1994. TechnischeUniversitatChemnitz{Zwickau,FakultatfurMathematik,1995. andr.g.voigt,editors,5thconferenceondomaindecompositionmethodsforpde, pages62{72,philadelphia,1992.siam. [16]P.LeTallec.DomainDecompositionMethodsinComputationalMechanics.ComputationalMechanicsAdvances(NorthHolland),1(2):121{220,Feb.1994. Review,34:581{613,1992. [17]J.Xu.Iterativemethodsbyspacedecompositionandsubspacecorrection.SIAM 9

OthertitlesintheSPCseries: 954M.Meyer.Grak-AusgabevomParallelrechnerfur3D-Gebiete.Januar1995 951T.Apel,G.Lube.AnisotropicmeshrenementinstabilizedGalerkinmethodsJanuar 952M.Meisel,A.Meyer.ImplementierungeinesparallelenvorkonditioniertenSchur-KomplementCG-VerfahrensindasProgrammpaketFEAP.Januar1995. 953S.V.Nepomnyaschikh.Optimalmultilevelextensionoperators.January1995 955T.Apel,G.Haase,A.Meyer,M.Pester.Parallelsolutionofniteelementequation systems:ecientinter-processorcommunication.februar1995 957M.Bollhofer,C.He,V.Mehrmann.ModiedblockJacobipreconditionersfortheconjugategradientmethod.PartI:Thepositivedenitcase.January1995 956U.Groh.EintechnologischesKonzeptzurErzeugungadaptiverhierarchischerNetzefur FEM-Schemata.Mai1995 9510G.Kunert.EinResiduenfehlerschatzerfuranisotropeTetraedernetzeundDreiecksnetze 959H.Matthes.ADDpreconditionerfortheclampedplateproblem.February1995 958P.Kunkel,V.Mehrmann,W.Rath,J.Weickert.GELDA:ASoftwarePackageforthe inderfinite-elemente-methode.marz1995 SolutionofGeneralLinearDierentialAlgebraicEquation.February1995 9513A.Meyer,D.Michael.Someremarksonthesimulationofelasto-plasticproblemson 9511M.Bollhofer.AlgebraicDomainDecomposition.March1995 9514B.Heinrich,S.Nicaise,B.Weber.Ellipticinterfaceproblemsinaxisymmetricdomains. 9512B.Nkemzi.PartielleFourierdekompositionfurdaslineareElastizitatsprobleminrotationssymmetrischenGebieten.Marz1995 parallelcomputers.march1995 9517C.He,A.J.Laub,V.Mehrmann.Placingplentyofpolesisprettypreposterous.May 9516W.Rath.Canonicalformsforlineardescriptorsystemswithvariablecoecients.May 9515B.Heinrich,B.Lang,B.Weber.ParallelcomputationofFourier-nite-elementapproxi- PartI:Singularfunctionsofnon-tensorialtype.April1995 9518J.J.Hench,C.He,V.Kucera,V.Mehrmann.DampeningcontrollersviaaRiccati mationsandsomeexperiments.may1995 9521A.Vogel.SolversforLameequationswithPoissonrationear0.5.June1995. 9519M.Meisel,A.Meyer.Kommunikationstechnologienbeimparallelenvorkonditionierten 9520G.Haase,T.Hommel,A.MeyerandM.Pester.BibliothekenzurEntwicklungparalleler Algorithmen.Juni1995. Schur-KomplementCG-Verfahren.Juni1995 equationapproach.may1995

9524P.Benner,R.Byers.Newton'smethodwithexactlinesearchforsolvingthealgebraic 9522P.Benner,A.J.Laub,V.Mehrmann.AcollectionofbenchmarkexamplesforthenumericalsolutionofalgebraicRiccatiequationsI:Continuous-timecase.October1995. 9523P.Benner,A.J.Laub,V.Mehrmann.AcollectionofbenchmarkexamplesforthenumericalsolutionofalgebraicRiccatiequationsII:Discrete-timecase.toappear:December Riccatiequation.October1995. 9526C.Israel.NETGEN69-EinhierarchischerparallelerNetzgenerator.August1995. 9527M.Jung.Parallelizationofmulti-gridmethodsbasedondomaindecompositionideas. 9525P.Kunkel,V.Mehrmann.LocalandGlobalInvariantsofLinearDierential-Algebraic EquationsandtheirRelation.July1995. 9529G.Windisch.Exactdiscretizationsoftwo-pointboundaryvalueproblems.October1995. 9530S.V.Nepomnyaschikh.Domaindecompositionandmultileveltechniquesforpreconditioningoperators.November1995. November1995. 9528P.Benner,H.Fabender.ArestartedsymplecticLanczosmethodfortheHamiltonian eigenvalueproblem.october1995. 9534Th.Apel,F.Milde,M.The.SPC-PMPo3D Programmer'smanual.December1995. 9532V.Mehrmann,H.Xu.Ananalysisofthepoleplacementproblem.I.Thesingleinput 9533Th.Apel.SPC-PMPo3D User'smanual.December1995. 9531H.Matthes.Parallelpreconditionersforplateproblems.November1995. 9535S.A.Ivanov,V.G.Korneev.Onthepreconditioninginthedomaindecompositiontechniqueforthep-versionniteelementmethod.PartI.December1995. case.november1995. 9536S.A.Ivanov,V.G.Korneev.Onthepreconditioninginthedomaindecompositiontechniqueforthep-versionniteelementmethod.PartII.December1995. directorypub/local/mathematik/spc.(notethecapitallinlocal!) Somepaperscanbeaccessedviaanonymousftpfromserverftp.tu-chemnitz.de, Thecompletelistofcurrentandformerpreprintsisavailablevia http://www.tu-chemnitz.de/~pester/sfb/spc95pr.html.